An S3 event triggers a pipeline that appends rows to a target. Duplicate notifications occasionally create duplicate loads. The bucket uses versioning, and each object version should be processed once at the business level. Which design changes are appropriate? Select 2 answers.

Published
  1. Use a stable processing identity based on the bucket, object key, and object version.
  2. Assume every S3 event notification is delivered exactly once and remove all deduplication.
  3. Mark every source version complete before validating or writing any target rows.
  4. Deduplicate permanently by the basename only, ignoring bucket, prefix, and version.
  5. Use an atomic claim or idempotent target commit with recoverable completion state, marking success only after the intended output is committed.

Comments — 0